功耗受限情况下多核处理器能效优化方案

发布时间：2018-01-09 04:08

本文关键词：功耗受限情况下多核处理器能效优化方案　出处：《计算机工程》2017年04期 　论文类型：期刊论文

【摘要】：将处理器功耗控制在预算以下有助于降低散热成本和提升系统稳定性,但现有功耗优化方案大多依赖线下分析得到的先验知识,影响实用性,而集中式搜索最优策略的算法也存在复杂度过高的问题。为此,提出功耗优化方案PPCM。利用动态电压频率调整(DVFS)技术控制CPU功耗在预算内以提高处理器能效。同时,将功耗控制和功耗分配解耦合以提高灵活性。采用动态调整的线性模型估计功耗,通过反馈控制技术对其进行调节。以计算访存比为指标在应用间分配功耗,并考虑多线程应用特征进行线程间功耗分配。实验结果表明,PPCM比Priority算法速度平均提高10.7%,能耗平均降低5.1%,能量-延迟积平均降低14.3%。与PCM CA算法相比,其速度平均提高4.5%,能量-延迟积平均降低5.0%。
[Abstract]:Controlling the processor power consumption below the budget can reduce the cost of heat dissipation and improve the stability of the system, but most of the existing power optimization schemes rely on the prior knowledge obtained from offline analysis, which affects the practicability. The algorithm of centralized search optimal strategy also has the problem of high complexity. In order to improve processor energy efficiency, a power optimization scheme, PPCM, is proposed. Dynamic voltage frequency modulation (DVFS) technology is used to control the power consumption of CPU within the budget. Power control and power allocation are decoupled to improve flexibility. The dynamic adjusted linear model is used to estimate the power consumption, and the feedback control technique is used to adjust the power consumption. The power consumption is distributed between applications with the index of computing access to memory ratio. The experimental results show that the speed of PPCM is 10. 7% higher than that of Priority algorithm, and the average energy consumption is reduced by 5.1%. Compared with the PCM CA algorithm, the average speed of the energy delay product is increased by 4.5 and the energy delay product is reduced by 5.0 percent.
【作者单位】：中国科学技术大学计算机科学与技术学院;
【基金】：国家“863”计划项目(2012AA010901)
【分类号】：TP332
【正文快照】： 中文引用格式:邱晓杰,安虹,陈俊仕,等.功耗受限情况下多核处理器能效优化方案[J].计算机工程,2017,43(4):39-45.英文引用格式:Qiu Xiaojie,An Hong,Chen Junshi,et al.Energy Efficiency Optimization Scheme for Multi-coreProcessor Under Pow er Consumption Constraint[J]

【相似文献】